I think the actuary-types who design the scoring models would disagree vehemently with this one. Scoring models are built from regressions on actual default rates.
It's one thing to say "I see you've been late on a few of your car payments, so we're going to give a higher interest rate." and it's something else to say "You have a few unpaid medical bills/late CC payments so we're going to give you a higher interest rate on this car loan."
I just have a hard time really believing that is the case. Often credit issuers simply look at the totality of everything, instead of appropriately looking the history of the specific financial contract they're wanting to issue.
That doesn't mean that systematic failures can't happen, but they're pretty unlikely. It's much more likely that you underestimate the correlation between these disparate defaults than that the banks overestimate it.
This quite literally happens already.